Nonlinear Feature-Based MI Detection Supported by DWT and EMD on ECG: A High-Performance Decision Support Approach

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Myocardial infarction (MI) presents a significant health risk with high mortality rates.
• Electrocardiogram (ECG) is a common diagnostic tool, but MI detection can be challenging due to subtle signal abnormalities and inter-observer variability.
• Advanced AI-based decision support systems are crucial for improving diagnostic accuracy in cardiovascular disorders.

Purpose of the Study:

• To develop and evaluate a robust AI framework for accurate myocardial infarction detection using ECG data.
• To enhance the diagnostic classification of MI by employing a hybrid feature extraction and selection methodology.
• To assess the potential of the proposed system for clinical decision support in cardiology.

Main Methods:

• Analysis of Lead II ECG derivations from healthy individuals and MI patients.
• Hybrid feature extraction using Empirical Mode Decomposition and Discrete Wavelet Transform to capture non-stationary signal characteristics.
• Extraction of 390 time, nonlinear, and complexity-based features using 23 entropy measures.
• Feature subset selection via Particle Swarm Optimization (PSO).
• Classification using Support Vector Machines, Artificial Neural Networks, k-Nearest Neighbors, and Bagged Trees.

Main Results:

• A hybrid feature representation combined with PSO-based selection significantly improved classification performance.
• The Bagged Trees classifier demonstrated the highest accuracy, achieving an overall correct classification rate of 97.6%.
• The optimized feature set proved effective in distinguishing between healthy individuals and MI patients.

Conclusions:

• The proposed hybrid feature engineering and PSO-based selection framework offers a reliable method for MI detection from ECG signals.
• This AI-driven approach shows strong potential for integration into clinical decision support systems, improving early MI diagnosis.
• The study highlights the effectiveness of advanced signal processing and machine learning techniques in cardiovascular diagnostics.